Skip to content

Make the front page "archive size" graphs consistent with one another

The archive size graphs on the frontpage of the archive are misleading in two ways:

  • the revision and origin graphs truncate the Y axis (so the bottom of the axis is not zero, which makes the increase look much larger than it actually is)
  • the revision and origin graphs don't have the same timescale as the content graph.

I'm not sure which way we should fix these issues (I guess @rdicosmo needs to make an executive decision here), but the current inconsistency is a common source of confusion.

The main problem to solve this issue is that we don't really have the old data for the revision counts and origin counts.

We don't even have proper data to "backfill" the graph for counts of revisions: while we initially recreated the graph for contents by using the ctime field on that table, we don't have such a field on other object types (and ctimes on contents are a pain, because their meaning across multiple mirrors of the archive is quite unclear, and they're not intrinsic, so we're unlikely to want to generalize them).

We could probably backfill the origin graph by using the date of first (successful) visit of any single origin.

And, in the end, we need to decide if we want these graphs to last "forever" or if a snapshot of the last year is sufficient.

Migrated from T2619 (view on Phabricator)